Preferably, in step (5), the segment is divided into m segments, for all data points in each code rate segment, data points with a confidence of 80% are screened, and the average code rate and the average score of the data points are obtained to obtain the operating point (Bitrate ') of the segment'_j，Score′_j) And m sections obtain m working points, n data points are set in a certain code rate interval, and then the sum of Euclidean distances between each data point and other data points is respectively obtained:

selecting 80% of data points with the minimum Euclidean distance sum according to the confidence coefficient of 80%; monotonously processing m working points, namely ensuring that the Score is high when the Bitrate is high; the line connecting the m operating points can be regarded as an RD curve.

Preferably, in step (6), the average code rate Bitrate required by the coding unit can be calculated from the RD curve_avgBitrate is used_avg1.5 as maximum bit Rate Bitrate_max(ii) a If the maximum code rate does not exceed the maximum code rate set by the user, coding at the code rate; otherwise, coding is carried out according to the code rate set by the user.

The invention has the beneficial effects that: the method has the advantages of greatly reducing the computational complexity, achieving the quality consistency when coding scenes with different complexities, being capable of adaptively determining coding parameters according to the quality scores set by a user, having lower computational complexity and being conveniently applied to the existing coding framework.

Drawings

FIG. 1 is a flow chart of the method of the present invention.

Detailed Description

The invention is further described with reference to the following figures and detailed description.

In the embodiment shown in fig. 1, a scene-based adaptive video coding method includes an analyzer and a predictor, where the analyzer is configured to determine a coding frame type and count coding information of each frame, where the coding information of each frame includes setting a fixed quantization parameter QP value, an actual consumed bit number, an actual quality score, and a skip block number; the predictor generates an RD curve of each scene according to the scene information and the coding statistical information, and outputs actual coding parameters according to user set parameters; the video encoder is a general h.264 or h.265 encoder; the specific operation steps are as follows:

the generated intermediate file is as follows: the set range of the fixed quantization parameter QP is [ QP_min，QP_max]For example, [20, 40 ] can be taken]The fixed quantization parameter QP of each frame is a fixed value; the fixed quantization parameter QP value of the I-frame is set to QP_minI.e. QP_I＝QP_minThe fixed quantization parameter QP value of the subsequent P frame is QP according to step size_stepMake an incremental setting (step size is typically 1) if the value is greater than QP_maxThen get back from QP_minStart setting, i.e. QP_Pidt＝QP_min+(idx*QP_step)/(QP_max-QP_min)，P_idxRefers to the idx-th P frame;

the specific calculation method is as follows: distributing the BIT number of the I frame to other P frames according to the proportion of skip blocks in each P frame in the same GOP, and setting the BIT number of the I frame as BIT_IThe BIT number consumed by the idx-th P frame is BIT_idxBIT number after correction is BIT'_idxCalculating to obtain a theoretical code rate by combining the frame rate FPS of the video file; then:

Bitrate_idx＝BIT′_idx*FPS

each P-frame in the GOP may be calculated to obtain a data point (bitate)_idx，Score_idx)；

(5) All data points (Bitrate) in the same scene are combined_idx，Score_idx) Sorting according to the code rate from small to large, processing in sections at certain code rate intervals, screening all data points in each code rate section, solving the average code rate and the average fraction of the data points, obtaining the working point of the section, and fitting an RD curve;

for example, the code rate interval is 200kbps, the interval is set to be m segments, data points with a confidence of 80% are screened for all data points in each code rate segment, and the average code rate and the average score of the data points are obtained to obtain the operating point (bite ') in the segment'_j，Score′_j) m sections obtain m working points, n data points are set in a certain code rate interval, and then the sum of Euclidean distances between each data point and other data points is respectively obtained:

selecting 80% of data points with the minimum Euclidean distance sum according to the confidence coefficient of 80%; monotonously processing m working points, namely ensuring that the Score is high when the Bitrate is high; the connecting line of the m working points can be used as an RD curve;

(6) according to the quality fraction set by a user, the predictor adaptively generates a coding rate for the coding unit;

specifically, the RD curve can be used to calculate the average bit rate (Bitrate) required by the coding unit_avgBitrate is used_avg1.5 as maximum bit Rate Bitrate_max(ii) a If the maximum code rate does not exceed the maximum code rate set by the user, coding at the code rate; otherwise, coding is carried out according to the code rate set by the user.

Claims

1. A self-adaptive video coding method based on scenes is characterized by comprising an analyzer and a predictor, wherein the analyzer is used for determining a coding frame type and counting coding information of each frame; the predictor generates an RD curve of each scene according to the scene information and the coding statistical information, and outputs actual coding parameters according to user set parameters; the specific operation steps are as follows:

(2) coding each coding unit according to the fixed GOP number, 0B frame, 1 reference frame and a fixed quantization parameter QP mode, and generating an intermediate file; the generated intermediate file is as follows: the set range of the fixed quantization parameter QP is [ QP_min，QP_max]The fixed quantization parameter QP of each frame is a fixed value; the fixed quantization parameter QP value of the I-frame is set to QP_minI.e. QP_I＝QP_minThe fixed quantization parameter QP value of the subsequent P frame is QP according to step size_stepMake an incremental setting if the value is greater than QP_maxThen get back from QP_minAt the beginning of the setup, i.e.

P_idxRefers to the idx-th P frame;

(4) Calculating theoretical codeRate, calculating each P frame in the GOP to obtain a data point; the specific calculation method is as follows: distributing the BIT number of the I frame to other P frames according to the proportion of skip blocks in each P frame in the same GOP, and setting the BIT number of the I frame as BIT_IThe BIT number consumed by the idx-th P frame is BIT_idxBIT number after correction is BIT'_idxCalculating to obtain a theoretical code rate by combining the frame rate FPS of the video file; then:

Bitrate_idx＝BIT′_idx*FPS

(5) Sorting all data points in the same scene from small to large according to code rates, processing the data points in sections at intervals of certain code rates, screening all the data points in each code rate section, solving the average code rate and the average fraction of the data points to obtain a working point of each code rate section, and fitting an RD curve;

2. The adaptive video coding method based on scenes as claimed in claim 1, wherein the coding information of each frame comprises setting a fixed quantization parameter QP value, an actual consumed bit number, an actual quality score, and a skip block number; the video encoder is a general h.264 or h.265 encoder.

3. The scene-based adaptive video coding method according to claim 1, wherein in step (5), the partition into m sections is setScreening data points with the confidence coefficient of 80% for all data points in each code rate section, and solving the average code rate and the average fraction of the data points to obtain the working point (bite ') of each code rate section'_j，Score′_j) M sections obtain m working points, n data points are set in a certain code rate section, and then the sum of Euclidean distances between each data point and other data points is respectively obtained:

4. The method as claimed in claim 1 or 2, wherein in step (6), the average bit rate (Birate) required by the coding unit is calculated from the RD curve_avgBitrate is used_avg1.5 as maximum bit Rate Bitrate_max(ii) a If the maximum code rate does not exceed the maximum code rate set by the user, coding at the code rate; otherwise, coding is carried out according to the code rate set by the user.